Short-term load forecasting (STLF) is crucial for energy management. Accurate and robust predictions remain challenging because load series exhibit non-stationarity, multi-scale dependencies, and stochastic variations. We propose ICEEMDAN-DAGT, a hybrid multi-component framework designed to tackle these complexities. The framework employs an improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) to decompose complex load signals. It incorporates an attention-enhanced dual-stream prediction module (DAGT), which uses a Gated Temporal Convolutional Network (TCN) to capture long-term dependencies and a Temporal-Feature Attention (TFA)-enhanced Gated Recurrent Unit (GRU) to model short-to-medium-term patterns. Dendritic Neuron Models (DNM) are used to strengthen nonlinear mapping, and an error-driven dynamic weighting strategy adaptively fuses subnetwork outputs. When validated on two real-world datasets (Panama City and Australia), ICEEMDAN-DAGT demonstrates superior performance that is statistically significant. On the Australian dataset, it achieves MAE reductions of 11.94 %-19.10 % and RMSE reductions of 12.48 %- 19.22 % compared to advanced baselines such as PatchTST and DLinear. These results confirm the framework's robust generalizability, improved accuracy, and computational efficiency, providing a promising solution for challenging STLF tasks.

Deep learning has shown substantial progress in fault diagnosis in recent years. Nevertheless, its application is hindered by excessive computational requirements and challenges in addressing data scarcity and multicondition adaptability in practical settings. Notably, the problem of feature redundancy stands out, as it not only inflates computational demands but also obscures vital features, reducing the generalizability and precision of diagnostic models. To address these challenges, this study proposes an efficient transfer fault diagnosis algorithm based on a multi-scale redundant feature selection network enhanced by parameter optimization. First, the proposed algorithm combines short-time Fourier transform (STFT) and synchrosqueezing wavelet transform (SWT) to comprehensively extract deep-level features from the data. Secondly, a channel and spatial redundancy feature selection unit is employed to effectively reduce channel and spatial redundancy in the features, thereby focusing on the learning of highly representative features. Furthermore, a transformer feature extraction (TFE) module is integrated at the backend of the model to accurately capture global contextual information and detailed features in the data. Finally, a multi-strategy improved Harris Hawk optimization algorithm (MS-HHO) was designed to achieve efficient hyperparameter tuning and global optimal configuration. Experimental results demonstrate that the proposed model achieved peak diagnostic accuracies of 98.63%, 98.83%, and 98.9% on three validation datasets using only 20% of the samples. Even under severe noise conditions, the model maintained an average diagnostic accuracy of 73.75%. Additionally, it exhibited superior diagnostic performance and robustness compared to competing models in diverse transfer tasks.

Currently, most mainstream gearbox fault diagnosis methods rely heavily on deep learning techniques. However, such deep learning-based diagnostic models often face challenges such as high computational resource requirements, poor real-time performance, and low diagnostic accuracy when applied to other operating conditions. To address the aforementioned issues, a novel lightweight cross-operating-condition gearbox transfer fault diagnosis framework is proposed, incorporating parameter optimization and adaptive feature attention enhancement. The framework utilizes two-dimensional Gramian angular field images to extract deep gearbox fault information. An improved lightweight mechanism integrating multi-scale channel and spatial adaptive attention enhancement is first adopted to allocate feature weights. This mechanism amplifies critical features while effectively mitigating the influence of redundant or noisy features. Subsequently, a local maximum mean discrepancy loss is employed to achieve cross-domain feature alignment across different operating conditions. Finally, a multi-strategy improved triangular topology optimization algorithm is applied to fine-tune the model's hyperparameters, boosting its cross-condition diagnostic capabilities and ensuring more accurate fault detection across different operational scenarios. Experimental results indicate that the proposed model achieved accuracy rates of 99.85% and 99.94% on two validation datasets, while maintaining diagnostic accuracies of 74.6% and 79.6% in the presence of significant noise. The model's diagnostic performance and robustness are superior to those of other models, demonstrating its effectiveness in challenging conditions.

Photovoltaic (PV) arrays' random and intermittent output characteristics impact power system safety. To improve the performance of the PV array fault diagnosis model, a novel online fault monitoring technique is introduced. (1) Fault diagnostic model construction: Significant differences in PV arrays' I-V and P-V curves under various fault conditions led to constructing a 3D channel feature map based on I, V, and P features. (2) Multi-source information fusion network (MSIFN): this multi-module fusion model includes a time-frequency domain fusion module (TDFM), a multi-feature shuffle expansion convolution module (MSECM), a parameterfree parallel hybrid attention enhancement module, and a multi-scale mixed pooling fusion classification module (MMPCM). (3) Multi-strategy fusion whale optimization algorithm (MSFWOA): addressing the original WOA's deficiencies, we designed time control, parameter modification, and greedy control strategies based on lens imaging to optimize MSIFN's hyper-parameters. Experimental results show that the MSFWOA-MSIFN model excels in PV array fault diagnosis (Paccuracy=Pprecision=Precall = 99.92 %). In three types of noise experiments with 15 dB, 25 dB, and 30 dB, the average performance index remained above 99 %. In practical experiments, the average performance indices werePaccuracy = 97.53 %, Pprecision = 97.32 %, andPrecall = 97.41 %, further demonstrating its excellent diagnostic performance. This model effectively diagnoses various faults in PV arrays, providing scientific and theoretical support for PV system operations.

Aiming at the problems of time-consuming and low diagnosis accuracy during the unsupervised training process of traditional DBN, an analog circuit fault diagnosis method based on Improved Multi-Objective Dragonfly Optimized Deep Belief Network (IMODA-ADBN) is proposed. The method employs an improved MODA algorithm instead of the BP algorithm, which improves the classification accuracy of the network and ameliorates the problem of being prone to falling into local optima. The algorithm is tested on three multi-objective mathematical benchmark problems and compared with three well-known meta-heuristic optimization algorithms such as MODA, MOPSO and NSGA-II, and the results demonstrate the stability of the IMODA-ADBN network model. Finally, IMODA-ADBN is applied to the diagnostic experiments of a two-stage quad op-amp dual second-order low-pass filter, and the results show that the method improves the classification accuracy and diagnostic rate while guaranteeing the convergence speed, and is able to effectively realize the classification and localization of difficult faults.

Photovoltaic (PV) arrays are installed outdoors and prone to abnormalities and various faults under harsh natural conditions, reducing power conversion efficiency and the life of the PV modules, and even causing electric shock and fire. Current fault diagnosis methods are unable to accurately identify and locate faults in PV arrays in PV power systems, leading to increased operation and maintenance costs. Therefore, the feature-enhancement improved dilated convolutional neural network (CNN) is proposed for fault diagnosis of PV arrays in this paper. Firstly, aim at the problem of information loss due to data structure and spatial hierarchy within the traditional CNN, and the loss of data after down-sampling, which leads to the inability to reconstruct information, a dilated convolution is introduced to obtain a larger perceptual field while reducing the computational effort. Meanwhile, the adaptive dual domain soft threshold group convolution attention module is proposed to enhance the essential features of faults and reduce the information redundancy given the ambiguity and blindness of the feature data in PV array fault extraction. Finally, the model performance of the proposed model is validated and the operability and effectiveness of the proposed method are verified experimentally. The diagnostic results show that the average diagnostic accuracy of the proposed model is 98.95% compared with other diagnostic models, with better diagnostic accuracy and more stable diagnostic performance.

In contemporary industrial processes, vibration signals collected from bearings often contain significant noise, challenging the efficacy of conventional predictive models in extracting critical degradation features and accurately predicting the remaining useful life (RUL) of bearings. Addressing these challenges, this paper introduces a novel method for predicting bearing RUL under noisy conditions, leveraging a dual-branch multi-scale convolutional attention network (DMCSA) integrated with a dense residual feature fusion network (DRF). Initially, the method applies continuous wavelet transform (CWT) to vibration signals to extract color time-frequency image data, followed by grayscale processing to construct a comprehensive color-grayscale time-frequency image dataset, thereby augmenting the model's input features. Enhanced channel and spatial attention mechanisms, combined with multi-scale convolutions, facilitate superior feature extraction and selection. The model's resilience to noise is fortified by incorporating noise into the training dataset. Subsequently, selected color-gray time-frequency features undergo fusion and relearning through the DRF framework at the model's backend. The crayfish optimization algorithm (COA) is deployed for the astute determination of the model's critical hyperparameters. The proposed DMCSA-DRF model is then applied to predict the health indicator (MSCA-DRF-HI) of the test dataset, culminating in the accurate prediction of the bearings' RUL. Validation experiments demonstrate that our method surpasses comparative models in prediction accuracy under diverse noise interferences, signifying a substantial advancement in predictive performance.

Research on ultra-short-term PV power prediction often overlooks complex spatio-temporal correlations and suffers from interpretability issues due to the black-box nature of deep learning. This study proposes an intelligent online prediction method to address these problems. The adaptive parallel spatio-temporal fusion network (APSTFNet) is introduced, integrating observation-aware and depth-aware modules for spatial feature capture and BiLSTM with self-attention for temporal dependencies. The dynamic adaptive weighted chimp optimization algorithm (DAWCHOA) optimizes APSTFNet’s hyperparameters. An interpretability framework uses the neuron conductance gradient method to elucidate the prediction mechanism. Experiments on a northern China PV power station show outstanding results: RMSE=1.12×10-1, MAE=0.89×10-1, MBE=0.15×10-1, R2=99.90. APSTFNet demonstrates superior robustness and stability, offering strong support for PV system operations.

To solve the problem of inadequate feature extraction and loss of key features in the process of remaining life prediction of bearings by traditional convolutional neural network (CNN), a CNN prediction model based on global attention mechanism (GAM) is proposed to achieve accurate remaining useful life (RUL) prediction of bearings. In this paper, a GAM–CNN prediction model for bearing RUL is proposed. First, the bearing’s one-dimensional (1D) vibration signal is transformed into two-dimensional (2D) image data that CNN is good at processing using continuous wavelet transform (CWT). Secondly, extract the time-domain degradation characteristics of bearings, select and build the health indicator (HI) of bearings using the monotonicity of the degradation characteristics, and introduce GAM into the CNN structure to adjust the contribution of key and unnecessary features to the bearing degradation process. GAM–CNN-based RUL prediction experiments were carried out with high-speed bearings of wind turbines and intelligent maintenance systems (IMS) experimental bearings, and the results show that GAM–CNN can effectively improve the prediction performance of the model. The results show that GAM–CNN has better prediction accuracy and generalization performance than other RUL prediction methods.

Photovoltaic (PV) arrays have output characteristics such as randomness and intermittency, and faults can seriously affect the safe operation of the power system. In order to improve the comprehensive performance of the PV array fault diagnosis model, a new intelligent online fault monitoring method for PV arrays is proposed in this paper. (1) a three-dimensional channel feature map based on I, V, and P features is constructed because the I-V and P curves of the PV array have significantly different effects under different fault conditions. (2) The PV array fault diagnosis model based on a multi-source information fusion network (MIFNet) is proposed, and Channel Mixing Convolution (CMC) module, three-dimensional feature attention enhancement (TDFAE) module, and Channel normalized scaling (CNS) module are designed to improve the comprehensive performance of the model. (3) An adaptive nonlinear mutual sparrow search algorithm (ANMSSA) is proposed to optimize the hyperparameter configuration of the MIFNet network. The experimental results show that the average recognition accuracy, prediction accuracy, and sensitivity of the ANMSSA-MIFNet network proposed in this paper are 99.64%, 99.64%, and 99.71% respectively. When facing single-component faults and multi-component faults, the model has stronger diagnostic accuracy, robustness, anti-noise ability, and stability, and can efficiently diagnose different faults of PV arrays, providing the scientific basis and theoretical support for the operation of PV systems.

In this study, the pulse rate (PR) of the artifact segments is predicted using a neural network. Predicting pulse rate be-comes a complicated problem due to the impaired collection of arterial blood pressure (ABP) signals. We propose to build a time series of interference pulse rate and discuss the best prediction model based on an Elman neural network under fixed and asynchronous length conditions. Meanwhile, the effectiveness and accuracy of the proposed method are validated when compared to the sliding window prediction method. The minimum RMSE, MAPE, and MAE of the proposed method were 0.0292, 1.9742%, and 0.0176. At the same time, this research laid the groundwork for discovering cardiovascular diseases.

To address the low fault feature extraction capability in analog circuits for component classification in analog circuits. Convolutional Block Attention Module-multiple-convolutional neural networks (CBAM-MIL-CNN) is proposed. The model has a better comprehensive performance in fault diagnosis experiments for circuits with secondary four-operator dual second-order low-pass filters, and can effectively achieve efficient classification and localization of all faults.

: Due to the low ability of fault feature extraction in analog circuits, it is impossible to classify components in analog circuits. A multi-input convolutional neural network (MIL-CNN) model based on attention mechanism is proposed. In the fault diagnosis experiment, the circuit of the two-stage four-op amplifier double-second order low-pass filter of the model has better comprehensive performance and can effectively realize the efficient classification and location of all faults.

Given the strong nonlinearity and large time-varying characteristics of membrane component fouling in the membrane water treatment process, a membrane component-membrane fouling diagnosis method based on the multi-objective jellyfish search adaptive deep belief network (MOJS-ADBN) is proposed. Firstly, the adaptive learning rate is introduced into the unsupervised pre-training phase of DBN to improve the convergence speed of the network. Secondly, the MOJS method is used to replace the gradient-based layer-by-layer weight fine-tuning method in traditional DBN to improve the ability of network feature extraction. At the same time, the convergence of the MOJS-ADBN learning process is proven by constructing the Lyapunov function. Finally, MOJS-ADBN is used in the membrane packaging diagnosis to verify the performance of the model diagnosis. The experimental results show that MOJS-ADBN has a fast convergence speed and a high diagnostic accuracy, and can provide a theoretical basis for membrane fouling diagnosis in the actual operation of membrane water treatment.

CBAM-MUL-CNN (convolutional block attention module - multiple - convolutional neural networks) model based on attention mechanism is proposed to solve the problem that the membrane fouling feature extraction capability of membrane bioreactor membrane component is insufficient, which resulted in the complex structure of the membrane fouling data, so that the efficient localization and classification of membrane fouling in membrane bioreactor could not be achieved. First, the time domain and frequency domain information about the fault data is used as the input of CNN (convolutional neural networks), and the features are extracted by convolution layer. Then, the input classifier is classified by splicing the time domain and frequency domain features using the full connection layer. BN (batch normalization) layer in the model can effectively prevent the disappearance of gradients, ReLU (rectified linear uint) layer can improve the non-linear model expression ability, CBAM (convolutional block attention module) can simplify the model complexity, improve the network features expression ability, and pooling layer can improve the model fault tolerance. The comparison results show that the model has excellent comprehensive performance in the membrane fouling diagnosis experiments of series tubular membrane devices and parallel hollow fiber membrane devices, and can effectively classify and locate all membrane fouling, making the treatment of water by membrane process improve the quality of effluent while reducing energy consumption, which provides a theoretical basis for actual production.

Compared to the traditional activated sludge process, the membrane bioreactor (MBR) has several advantages such as the production of high-quality effluent, generation of low excess sludge, smaller footprint requirements, and ease of automatic control of processes. The MBR has a broader prospect of its applications in wastewater treatment and reuse. However, membrane fouling is the biggest obstacle for its wider application. This paper reviews the techniques available to predict fouling in MBR, discusses the problems associated with predicting fouling status using artificial neural networks and mathematical models, summarizes the current state of fouling prediction techniques, and looks into the trends in their development.

The difficulty in extracting the fault features of analog circuit leads to complex calculation and poor precision with the model. A fault diagnosis method for analog circuits based on attention mechanism and convolutional neural network (CBAM -CNN) is proposed. Firstly, the image features of the input layer were extracted by using the convolution kernel. Followed by rectifying linear unit (ReLU) was connected behind each convolution layer, and a batch normalization (BN) layer was added to solve the problem of internal covariate migration, so as to improve the expression ability of the nonlinear model. Secondly, the convolutional block attention module (CBAM) was added after the batch normalization layer to extract the important features. After CBAM, the pooling layer is connected to reduce the computational complexity of the network and improve the accuracy and efficiency of the network. Finally, the Sallen-Key low-pass filter and the two-stage four-op amplifier double-order low-pass filter are taken as the research objects. The results of fault diagnosis experiments demonstrate that the proposed method can effectively improve the diagnosis accuracy and realize the classification and location of all faults with high difficulty.